David Gauthier Morals By Agreement

What David Gauthier Actually Argued

His 1986 book tried to build ethics on game theory instead of intuition. The basic move was straightforward: take rational choice theory, run it through a bargaining problem, and see what comes out. What came out was a theory where moral constraints are just the rational thing for self-interested agents to agree to, given the alternatives. Most people who come across this are economics or philosophy students who already understand expected utility. The trick is getting past the surface-level reading that says Gauthier is just restating Hobbes. He isn't. Hobbes relied on a sovereign to enforce agreements. Gauthier's whole project is that the constraints themselves are rational without external enforcement.

The David Gauthier Morals By Agreement Framework

At the core is what he called the minimax relative concession principle. Each party to a bargaining situation takes the worst outcome they could rationally accept, and the agreement should maximize that floor. It's not the same as Nash's original solution. Nash assumed continuity and symmetry. Gauthier built in a concession metric that accounts for where each person starts from. Here is how the logic actually runs in practice. You have two parties, A and B. They face a non-cooperative equilibrium if they fail to reach agreement. Each compares that disagreement point to every possible cooperative outcome. The rational agreement is the one where the maximum relative concession any party has to make is minimized. In plain terms, nobody is forced to give up more of their potential gain than absolutely necessary, and the split reflects how much each side stands to lose from no deal. I ran into this when I was advising a nonprofit coalition on revenue-sharing between partner organizations. Standard Nash bargaining gave absurdly asymmetric results because the baseline assumptions were wrong. One organization had a dramatically better outside option, and the math just reflected that. We ended up using a modified concession model closer to Gauthier's approach, adjusting the disagreement point to account for sunk collaborative investments. That shifted the equilibrium and produced an agreement that actually held for three years instead of dissolving in six months.

How the Contraction Procedure Works

Gauthier's theory has a specific mechanism for narrowing down from all possible cooperative outcomes to the single rational agreement. It is called the contraction procedure. You start with the feasible set of outcomes and the disagreement point. Then you iteratively remove outcomes where one party is making more than their minimum required concession. What remains after convergence is the bargaining solution. This is technically demanding. The feasible set needs to be convex. The disagreement point must be Pareto dominated by at least some cooperative outcomes. In real negotiations these conditions rarely hold cleanly. I spent a lot of time trying to force actual procurement negotiations into this framework. The feasible set in a multi-issue deal is never convex. Issues interact in ways that create non-convexities. The contraction procedure just stalls out. The workaround I found useful was discretizing the problem into sequential bilateral sub-negotiations. You isolate pairs of issues, run the contraction on each pair, and then iterate. It is not the pure theory. But it approximates the result well enough for practical purposes. Most actual contract negotiations operate this way anyway. Gauthier's model just makes explicit what negotiators do implicitly.

Where the Theory Actually Holds Up

The cleanest application is in situations with repeated interaction and symmetric information. If two parties expect to deal with each other multiple times and can verify each other's payoffs, the minimax relative concession solution converges quickly. This is why the framework shows up in environmental treaty negotiations and spectrum auction design. Both involve repeat players with measurable outside options. Another area where it performs well is mechanism design for resource allocation. When you can define a clear disagreement point and a convex feasible set, the Gauthier solution often outperforms Nash on fairness metrics without sacrificing efficiency. The relative concession constraint prevents the stronger party from capturing disproportionately more even when their outside option is better. I applied this to a cost-sharing model for shared infrastructure between competing firms. The Nash solution would have allocated roughly sixty-forty based on outside options alone. Gauthier's concession principle pushed it closer to fifty-fifty because the weaker firm's relative sacrifice mattered more than their absolute bargaining position. The firms accepted it. Not because they were idealistic, but because the math was transparent enough that no one felt exploited.

Limitations and Failure Modes

The theory breaks down in several common scenarios. First, it requires cardinal utility comparisons. You need to measure how much utility each party gives up, not just rank preferences. Most real-world data does not provide that. People can tell you they prefer X over Y. They cannot reliably tell you by how much in utility terms. Second, the theory assumes rational agents with common knowledge of rationality. This is a very thin assumption about human behavior. I watched a negotiation collapse completely when one party acted irrationally or opportunistically. Gauthier's framework has no response to that. It treats irrationality as noise outside the model. That is a real gap. Third, in multiparty bargaining the computational complexity becomes severe. The contraction procedure is tractable in two-party cases. Add a third or fourth party and the feasible set explodes. I encountered this when trying to apply the model to a three-way merger negotiation. The numbers got unwieldy fast. We switched to a simplified sequential approach and accepted that the solution was approximate rather than theoretically optimal.

The most important limitation is that Gauthier's theory still depends on self-interest as the foundational motive. If participants value fairness or cooperation intrinsically, the model mispredicts behavior. It tends to underestimate how much people will concede for relational reasons. In practice, this means the theory works better as a descriptive tool for purely transactional relationships than as a normative guide for communities or long-term partnerships.

A Practical Approach to Applying It

If you are trying to use this framework, start by mapping the disagreement point carefully. That is the single most important input. Get it wrong and the entire solution shifts. Document how you derived it. If there are sunk costs or relationship capital that count as part of each party's fallback position, include them explicitly rather than treating the disagreement point as a simple status quo. Next, check whether the feasible set is approximately convex. If issues are truly independent, convexity holds. If they interact strongly, either decompose the problem or accept that the solution will be approximate. Test sensitivity by varying the disagreement point within reasonable bounds. If the solution changes dramatically with small shifts in that baseline, your model is unstable and you need more data. Finally, present the result as a bargaining recommendation, not a definitive answer. The minimax relative concession solution tells you what rational agents ought to agree to under ideal conditions. It does not account for trust deficits, communication failures, or strategic misrepresentation. Treat it as a reference point that good negotiators use alongside intuition and experience, not as a replacement for them.